The longer I am around the agriculture industry, it seems the more interested I become in seeing what happens next.
In a column that appeared in early March, I talked about when I arrived on the scene of the family farm in eastern Ontario in the 1950s. It was at a time when real or literal horse power had quite recently been phased out and tractor power had taken over most, if not all, yard chores and field operations.
Well let’s rocket ahead a few light years and in this column written in late February I learned about a Vancouver-based company (with strong ties to Alberta) that has developed technology and software that can do facial recognition on your beef or dairy cowherd (horse, pig and sheep technology is in the works), which can help farmers monitor those animals on a daily basis and provide a wide range of biological information to help producers with all kinds of management decision making.
Let’s step back a second — yes facial recognition on cattle. A Holstein is a Holstein, a Red Angus is a Red Angus, a black baldie is a black baldie. Who knew that facial recognition could work on livestock? Well, Mokah and Geoffrey Shmigelsky of OneCup AI (that AI stands for artificial intelligence) are here to tell you that, yes, it can be done and it works.
If anyone has watched TV crime fighters in the past couple of years, you’ll know that at least on the TV set facial recognition is widely, quickly and confidently used to identify and catch the bad guys. (Did you ever notice how fast their technology works? I would still be trying to get my laptop turned on and opened up to a desktop in the time it takes them to capture a grainy cellphone image, scan it for facial recognition, identify the suspect, read a lengthy criminal record, get a home address, present it all in a nice PowerPoint presentation so everyone in the room can see, and dispatch the ace crime-fighting team to smash down the suspect’s door.)
Human facial recognition is a similar technology concept that has been developed by the Shmigelskys and their team to work in identifying individual animals in a beef or dairy herd. In fact, animal facial recognition is even more detailed than human facial recognition, says Mokah Shmigelsky.
With human facial recognition, the software connects with 128 parameters or reference points on the human face. With livestock, the software identifies 512 parameters or reference points on the cow’s face. If you had an all black, all red or all white cow face, the software can distinguish from the next animal with an all black, all red or all white face. This software isn’t just looking at a distinctive colour pattern in the hide, it’s measuring lips and counting eyelashes too.
Mokah, who was born and raised around the family beef business near Airdrie, just north of Calgary, says the idea for this whole animal facial recognition project stemmed from comments made by family beef producers around a campfire one summer day who mused, “Wouldn’t it be nice if there was some way photo technology could be used to monitor and measure livestock performance during the animal’s life on the farm.” Talk about throwing out a challenge.
That casual campfire suggestion got Mokah thinking about ways to automate many of the manual decision-making processes a rancher faces every day. It became the driving force behind OneCup AI.
And, of course, being teamed up with husband, Geoffrey Shmigelsky, was a great resource. He’d had great success in leveraging cutting-edge technologies and launching several start-ups. He holds a master’s degree in computer science from Georgia Tech, with a specialization in artificial intelligence and deep learning. (Personally, I am all over that deep learning thing, as soon as I find out what it is, but don’t wait for me.)
I can’t say I knew Geoffrey Shmigelsky, but we had crossed technology paths years earlier. About 30 years ago, he launched one of the first internet services in Calgary called CadVision, of which I was an early subscriber, and he offered workshops to explain to people how the internet worked. Eventually he sold CadVision to Telus.
But let’s get back to the real point of this cowherd facial recognition. The technology developed by OneCup AI and referred to as BETSY will not only use cameras and software to identify individual animals in the herd but will be capable of so much more.
While some elements are still in the development and testing stage, the technology will be there to allow BETSY to track animal health, animal activity, animal nutrition and animal growth all by using cameras to monitor individual animals daily and provide reports to the farm computer or cell phone.
Mokah Shmigelsky explains the basic package from OneCup AI will include a computer along with four to six cameras, all capable of communicating with each other through Wi-Fi. Their service also offers a Wi-Fi upgrade or booster kit so service is expanded and strengthened over the farm.
The cameras are to be mounted in the barn, in the yard or a nearby calving pasture, or if it is grazing season, even out on pasture. The idea is to place the cameras where cattle will pass or congregate — near a water source, a mineral tub, or a gateway through which they pass.
Ideally, it will work best for a camera to be able to see each animal at least once every two or three days. This is particularly important for calves or young stock that are still growing. Each time the camera sees newborn calf #45689, for example, it will update its facial recognition as the calf’s looks change and mature.
Again, aside from the facial recognition software, BETSY will eventually include something like infrared technology able to track diseases or symptoms such as fever, respiration, coughing, limping, arched back and depression.
Software will also be able to monitor each animal’s activity, reporting time spent running, walking, laying down, sleeping and socializing, along with breeding patterns. It will be able to track a calf’s activity to show whether it is nursing properly and bonding.
BETSY will also be equipped with nutrition software to measure feed intake and output. And there will also be software to track animal growth and weight.
While BETSY software will assign each animal an identification number, the software will also allow farm I.D. numbers from dangle tags and even CCIA RFID tag numbers to all be linked together for each animal.
As BETSY is starting out with cattle, Shmigelsky says plans are to expand software to work with horses, sheep and pigs.
At this column appears, BETSY is in beta testing with a number of producers and OneCup AI plans to offer BETSY to about 100 or more early adopters later this spring, just to give it all a good field trial before there is a full commercial rollout. OneCup AI is also working with several industry partners including Lakeland College, Olds College, Agriculture and Agri-Food Canada, and NVIDIA Inception Program (those are the deep learning people).
And in the meantime, if you have any trouble getting your tablet turned on just let me know. So far, I have about an 85 per cent success rate.